Xilinx is aming to promote wider use of its ‘all-programmable’ devices, for example it Zynq series of parts, although it says that the package is also applicable to its FPGA product lines. The application spaces it has in mind feature concepts such as the cobot – collaborative robots, where intelligent vision systems, possibly combined with other inputs from sensor fusion, will be required for safe operation of machines working is close proximity to people. Other product areas include ‘sense and avoid’ drones, augmented reality, autonomous vehicles, automated surveillance and medical diagnostics.
reVISION is intended to enables a much broader set of software and systems engineers, with little or no hardware design expertise, to develop intelligent vision guided systems easier and faster, combining machine learning, computer vision, sensor fusion, and connectivity. Systems must be extremely responsive, and the latest algorithms and sensors need to be quickly deployed.
reVISION enables, Xilinx claims, the fastest path to the most responsive vision systems, with up to 6x better images/second/watt in machine learning inference, 40x better frames/second/watt of computer vision processing, and 1/5th the latency over competing embedded GPUs and typical SoCs. Developers with limited hardware expertise can use a C/C++/OpenCL development flow with industry-standard frameworks and libraries such as Caffe and OpenCV to develop embedded vision applications on a single Zynq SoC or MPSoC.
Developers can use the stack to rapidly develop and deploy upgrades. Reconfigurability is critical to ‘future proof’ intelligent vision-based systems as neural networks, algorithms, sensor technologies and interface standards continue to evolve at an accelerated pace. The reVISION stack includes support for the most popular neural networks including AlexNet, GoogLeNet, SqueezeNet, SSD, and FCN. The stack provides library elements including pre-defined and optimized implementations for CNN network layers, required to build custom neural networks (DNN/CNN). The machine learning elements are complemented by a set of acceleration-ready OpenCV functions for computer vision processing.
For application level development, Xilinx supports industry-standard frameworks including Caffe for machine learning and OpenVX for computer vision. The reVISION stack also includes development platforms from Xilinx and third parties, including various types of sensors.
“Our eye tracking technology has greatly benefited individuals suffering from ALS or other forms of paralysis, powered by the Zynq SoCs for high resolution vision based analytics,” said Robert Chappell, CEO and Founder, Eyetech Digital Systems. “The new reVISION stack offers new opportunities for algorithm development by leveraging the power of machine learning. This can enable us to expand our offering of human interaction hardware as well as improving our core eye tracking products.”
“We are seeing tremendous interest in machine learning from the edge to the cloud, and believe that our ongoing investment in development stacks will accelerate mainstream adoption.” said Steve Glaser, SVP of Corporate Strategy at Xilinx. Glaser adds that the system includes support for the multiple stages of implementation and design refinement that will characterise machine-learning projects; the ‘training’ aspect of a project might be carried out in a floating-point math context, on GPUs – and then be implemented for deployment in fixed-point, on a programmable logic platform, for minimum latency.
The reVISION stack will be available in the second quarter of 2017; www.xilinx.com/revision